关键词: 文章发布时间检测, 网络信息挖掘, 网页内容分析, 文本消重

Abstract: Determination of the earliest time when an event is reported on the Web is of particular use for computer aided social science researches. Statistics has shown that 40% of Web pages have no evidence of publication time from their contents. For those cases， the crawling time or LMT （last-modified-time） from the HTTP header are often far off the real publication time. Therefore two methods for achieving better accuracy are proposed. The first one is based on link analysis and the other is based on replicas analysis. Experiments have shown that combining these two methods often gives rise to quite accurate results.

Key words: publication time, information mining, content analysis, replica detection